Inevitable Futures in Medicine | Eli Ben-Joseph (Regard)

A vision into the future of medicine and how Regard is helping to create it

This week I have the pleasure of writing about a recent conversation with the CEO and co-founder of Regard, Eli Ben-Joseph. Our conversation covered his journey starting Regard, his vision for the future of medicine, the promises and perils of LLMs in a healthcare setting, and more. Eli is a rare combination of humble yet incredibly ambitious founder; I am grateful he is dedicating his life to improving patient outcomes through the work that Regard is doing every day.

Eli’s Journey To Starting Regard

Eli's journey into the heart of healthcare innovation wasn't born out of a long-standing desire to become a physician. Instead, it was an unwavering interest in the intersection of computers and biology that led him down an unconventional path. With an undergraduate bioengineering degree from MIT and a graduate computer science / machine learning focused degree from Stanford, Eli knew that he wanted to put his interest in the two fields together, but at the time it wasn’t yet obvious how. In 2016, after graduating from Stanford during the first AI “deep learning” hype cycle and re-connecting with his now co-founder Nate, Eli & Nate spent 6 months figuring out how they could apply advancements in AI / machine learning to healthcare. They were eventually admitted to Techstars’ Cedars-Sinai accelerator, giving them an inside view into how hospitals and doctors operated day-to-day.

During this time, Eli & Nate recognized the shocking amount of time that physicians spent doing administrative work on a computer. According to Eli, when he shadowed doctors at the hospital, they were spending 15 minutes on the computer for every 1 minute they were with a patient. Eli also noticed that the healthcare system had a knack for continually adding information to a patient’s Electronic Medical Record (EMR), leading to a deluge of data that doctors didn’t have time to thoroughly sift through. Through more shadowing and conversations with physicians, it was increasingly clear that the burden of administrative tasks and data overload was leading to physicians’ disillusionment with their jobs and burnout. Physicians went through 8+ years of training to become doctors; they did not do it to take notes, write letters, and sift through endless amounts of data in a patient’s EMR.

These 6 months spent in the Cedars-Sinai accelerator allowed them to see firsthand the dissatisfaction among doctors and the inefficiencies plaguing the healthcare system. This realization was the inception of Regard: a company aimed at using AI to address the deluge of data and monotonous tasks overwhelming doctors, leading to physician burnout, suboptimal patient treatment, and hospital losses due to missed diagnoses and insurance denials.

Before Regard and After Regard

Regard's mission is simple yet profoundly ambitious: Bring world-class healthcare to everyone. They do this by employing AI and data science to streamline the input of patient data into systems and by distilling vast amounts of medical information into high-signal, actionable insights for doctors and patients.

Eli bolstered this vision with an apt analogy - much like how Language Models (LLMs) make the vast internet navigable by compressing information, Regard endeavors to compress the patient medical record alongside general medical knowledge, making high-value information readily available. This enables doctors to focus on delivering superior patient care, thereby improving patient outcomes and hospital operations.

Inevitable Futures

During our conversation, Eli spoke often about a concept he calls “inevitable futures”. He said that he and his co-founder Nate use it as a framework to think about where Regard is headed as a business. Much in the same way that electric vehicles and renewable energy are inevitable futures for transportation and energy creation, Eli sees the inevitable future in healthcare and is building Regard one day at a time to bring that future forward.

I love this idea, not only for its ambition but also because it implicitly highlights that, while these futures may be inevitable, they won’t happen on their own. It is therefore up to entrepreneurs and builders to see the future, and then step-by-step go build it.

In this future, interactions with healthcare will be profoundly transformed. Eli asked me to “Imagine an AI doctor, powered by Regard, distributed through healthcare systems, that acts as your initial medical consultation. Through your phone, this AI doctor assesses your symptoms, directs you to the appropriate lab tests if needed, and provides a personalized treatment plan—all without stepping foot in a doctor's office. For routine concerns, this is no doubt the future of healthcare. Of course, there will be times when you will still need to go see a doctor in person.”

To make this vision a reality, Regard is laying the groundwork by building a viable business that offers tangible value to doctors, patients, and healthcare systems today.

Regard, today

Regard’s Trajectory

My conversation with Eli and the task of writing this post afterward got me thinking about Maslow’s hierarchy of needs and whether a similar pyramid framework could be applied to healthcare. I was hoping that this would help me better understand where Regard exists today vs. the inevitable future it is creating. The result of that thought exercise can be found below.

Recalling the origin story of Regard, we know that doctors today are spending a lot of time on the lowest rungs of the pyramid, and not enough time on the rungs above. That’s what Eli identified when he started Regard in 2016, and it’s the value they provide for doctors and hospitals today (black brackets) - from helping to input patient data into the EMR through to recommending diagnoses.

In the future, Regard will help with up to the 4th rung of the pyramid, the Treatment plan, therefore enabling doctors to spend more time on the top rung, Treatment, while validating that everything below has been done properly by their Regard AI assistant. In addition, Eli mentioned that Treatment generates more data, creating a positive feedback loop from Treatment to Patient data as indicated in the diagram above. Eli mentioned that Regard can then apply AI and data science to this massive amount of patient data, helping to advance medicine at a much more rapid pace than ever before. Yet another opportunity for Regard in the future.

What About LLMs?

While Regard was founded in 2016, well before the popularity of LLMs, it is now finding clear use cases for LLMs in its business. Eli and I discussed at length how LLMs were being applied to healthcare, which was highly educational for me and shone a light on the promises and perils of applying this technology where human lives are on the line.

The Promises of LLMs in Healthcare

The promises of LLMs in healthcare are starting to become clear, both in terms of current capabilities and future ones.

LLMs already aid in various non-clinical tasks today, such as automating sick notes, discharge summaries, and responses to letters of denial from insurers. Through Regard’s partnership with OpenAI, they’ve shipped an LLM-powered AI assistant called Max, which assists physicians with non-clinical tasks today. In the future, powerful AI will inevitably assist with clinical tasks, though we are not currently there due to the perils mentioned below.

I asked Eli how Regard uses the OpenAI API on top of the patient’s EMR to provide more relevant medical analyses. Eli mentioned first and foremost that the OpenAI service they use is HIPAA compliant. Next, he added that they use Retrieval-Augmented Generation (RAG) to provide the general purpose, foundational models with relevant patient data from their EMR. According to Eli, RAG leads to fewer hallucinations and more relevant assessments of patients’ medical situations. Regard is now leveraging this proven technique to deliver value to its users. Eli mentioned that in the future, if a patient’s entire medical record can fit inside of a foundational model’s context window, this might remove the need to do RAG, but for the time being, RAG is a clear necessity.

Furthermore, model advancements like GPT-5, rumored to be far more powerful than GPT-4, point to a future where LLMs can reliably be used in a healthcare setting and potentially even make novel discoveries to improve human health. However, time and lots of testing will determine how reliable these models can be when patients’ lives are at risk.

The Perils of LLMs in Healthcare

The perils, however, are not to be underestimated. The accuracy required in medicine is unparalleled (Eli mentioned proudly that Regard’s diagnostic algorithms boast 98-99% acceptance rates by doctors), and any instance of hallucination—where AI generates incorrect or fabricated information—could have dire consequences (e.g. loss of human life). The challenge lies in enhancing the accuracy of these AI systems and establishing robust safeguards against misinformation.

For example, for an AI system to not require FDA approval, the algorithm must be explainable; black-box algorithms require FDA approval. Given that LLMs behave in a black-box way today, they will have to undergo rigorous scrutiny by the FDA before they can be used in a clinical setting, thus limiting their potential for more immediate impact.

There is no doubt that these systems will get better as time progresses, but today, there are very legitimate reasons that these systems are not “let loose” in clinical settings.

Reflections on Building in Healthcare

Reflecting on roughly 7 years building in healthcare, Eli shared a few final insights from his journey:

1. The extraordinary patience required to innovate in healthcare

2. The peculiarities of navigating healthcare systems and the consensus-driven decision-making that often slows progress

Yet these challenges have not deterred him.

Regard's unique position, with direct access to hospital patient records, sets it apart, offering a significant data advantage that could drive unparalleled innovations in healthcare.

At the end of our call, I asked Eli what I ask each Steelhead interviewee, which is what his Steelhead goal in life is - a metaphor for a challenge that is immensely difficult but profoundly rewarding. Eli said:

What I’m doing with Regard. I chose to go to Stanford to have massive impact on the world, that’s what I’m most excited about. I believe that the largest impact I can have is in health and healthcare; something that can have a step-change impact on the way people receive healthcare. My work with Regard is aligned with that, and while it comes with a high chance of failure and is really hard, if it works it could have a massive impact on the world - not just on the people in health systems in this country, but on anyone anywhere in the world with a computing device and the internet.

Thanks so much for using your skills to create an inevitable future that will benefit humanity, Eli.

-Erik